> I fail to see how any sort of much lower level access to the machine would be portable across price ranges and microarchitecture generations

That's the assumption that can be removed. GPUs don't have stable ISAs, and their assembly-like code gets recompiled for each microarchitecture.

In the Intel's world of prebaked machine code adoption of a wider set of SIMD instructions takes a decade+. In GPUs it's just a driver update.

Sure, but SIMD extensions and base ISA solve different problems.

As for GPUs, while the ISA is not constant, it's STILL C-ish running over a dynamic hardware scheduling layer.

Edit: To clarify, I have nothing against a closed ISA, I just don't see how making that ISA non-C-ish is valuable.